Reproduce “Dimension-Free Private Mean Estimation is Impossible Without the Covariance” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Renaud Gaucher, Ephraim Linder, Adam Smith, Jonathan Ullman, Lydia Zakynthinou
The authors study differentially private estimation of a Gaussian population mean when its covariance is unknown. They prove lower bounds under two constructed covariance priors: a random low-rank subspace and a diagonal heavy-tailed model. Under specified sample-size and privacy conditions, unknown covariance causes dimension-dependent error relative to known covariance; these are prior-averaged theoretical bounds, not universal claims.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.